Method of adaptive pixel averaging for impulse noise reduction in digital images
Adam Konieczka, Julian Balcerek, Adam Dąbrowski · 2018
In this paper a new nonlinear method for salt and pepper and random-valued impulse noise reduction in digital images is proposed. The authors achieved high effectiveness combined with a low computational complexity. The proposed solution is based on iterative process of noisy pixels detection using noise profiles and replacement of these pixels using the average values of neighboring noise free pixels. Noise profiles are trained using artificially noised test images. Experiments were performed on standard test images in order to compare proposed approach with other popular techniques. A quality measurement was the peak signal-to-noise ratio (PSNR). In many instances the proposed method was better than other existing techniques of impulse noise reduction.